• Phonetics and Speech Sciences
• /
• v.7 no.4
• /
• pp.9-16
• /
• 2015

In this paper, several experiments are performed according to deep neural network (DNN) based features for the performance comparison of speaker verification (SV) systems. To this end, input features for a DNN, such as mel-frequency cepstral coefficient (MFCC), linear-frequency cepstral coefficient (LFCC), and perceptual linear prediction (PLP), are first compared in a view of the SV performance. After that, the effect of a DNN training method and a structure of hidden layers of DNNs on the SV performance is investigated depending on the type of features. The performance of an SV system is then evaluated on the basis of I-vector or probabilistic linear discriminant analysis (PLDA) scoring method. It is shown from SV experiments that a tandem feature of DNN bottleneck feature and MFCC feature gives the best performance when DNNs are configured using a rectangular type of hidden layers and trained with a supervised training method.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1990.04a
• /
• pp.299-311
• /
• 1990

A computer simulation model is proposed for selecting effective agrees paths and estimating the evacuation time in case of evacuating all the residents of a building to the safe area. The physical structure of a building is formulated into a network. The followings are considered in the model : the congestion effect to the velocity, the behavioral pattern of the evacuees, and the probabilistic nature of the evacuation process. Coded in SLAM II/PC, the simulation model can generate the output such as evacuation time(max, average), utilization of exits, utilization of passages, floor clearance times, and bottleneck information. The "rigorous" validation of the proposed model is not completed yet but remains to be a future research topic.rch topic.

• Proceedings of the KIEE Conference
• /
• 2002.11c
• /
• pp.27-30
• /
• 2002

In this paper, we propose a window-based congestion control algorithm to achieve a fair sharing of the available bandwidth in ECN capable heterogeneous TCP networks with a single bottleneck link. The proposed algorithm is based on extracting the network status from the successive binary congestion information provided by ECN. From the explicit network information, we estimate the fair window size proportional to the propagation delay. Through simulations, the effect and performance of the proposed algorithm are shown for the heterogeneous networks.

• Journal of the Korean Society for Precision Engineering
• /
• v.20 no.4
• /
• pp.174-182
• /
• 2003

Recently, medical imaging has taken interest on CAD based solution for anatomical part fabrication or finite element analysis of human body. In principle, contours representing object boundary are obtained through image processing techniques. Surface models are then approximated by a skinning method. For this, various methods should be applied to medical images and contours. The major bottleneck of the reconstruction is to remove shape inconsistency between contours and to generate the branching surface. In order to solve these problems, bi-directional smoothing and the composite contour generation method are proposed. Bi-directional smoothing has advantage of removing the shape inconsistency between contours and minimizing shrinkage effect with a large number of iterations. The composite contour by the proposed method ensures smooth transition in branching region.

• Journal of the Korea Society for Simulation
• /
• v.6 no.2
• /
• pp.89-104
• /
• 1997

Performance bottleneck of parallel computer systems has mostly been I/O devices because of disparity between processor speed and I/O speed. Therefore I/O node placement strategy is required such that it can minimize the number of I/O nodes, I/O access time and I/O traffic in an interconnection network. In this paper, we propose an optimal distance-k embedding algorithm, and analyze its effect on system performance when this algorithm is applied to n x n torus architecture. We prove this algorithm is an efficient I/O node placement using software simulation. I/O node placement using the proposed algorithm shows the highest performance among other I/O node placements in all cases. It is because locations of I/O nodes are uniformly distributed in the whole network, resulting in reduced traffic in the intE'rconnection network.

• Proceedings of the Safety Management and Science Conference
• /
• 2004.11a
• /
• pp.47-51
• /
• 2004

Facility layout for enterprise's competitive power strengthening and construction of offensive manufacture line and direction presentation for design of production management system that is new by operation improvement according to S company. In enterprise in JIT(Just in Time) system of leading persons affecting in productions among a lot of habitat factor analyze. It is that through equipment Layout among those factors enterprise my problem improve and serves enterprise's absolute ability cultivation. Even if it is JIT production main point that affect to apply the modular production to productive system among at equipment Layout, lot size, bottleneck session etc. worker's multi-function anger, change of demand factor and productive capacity add. 'S' Through enterprise's example, I wish to establish Korean medium and small enterprises' manufacturing industry style equipment Layout model by taken place effect interpretation.

• Current Optics and Photonics
• /
• v.4 no.6
• /
• pp.545-550
• /
• 2020

We propose a hole-filling method to solve discontinuous depth representation and to reduce the visible seams and cracks that cause the limitation of the viewing angle of the three-dimensional (3D) image in the multilayer type 3D display system. The occlusion and the disocclusion regions between layers, such as the visible seams and cracks, are a major bottleneck of the multilayer type 3D display system to represent a volumetric 3D image by stacking multiple images. As a result, in the reconstructed 3D image, the visible seams and cracks appear as brighter overlapping and undesirable cut-off. In order to resolve the problems above, we applied the depth-fused effect to the sub-depth map generating algorithm and improve the viewing characteristics of the multilayer type 3D display. The experimental demonstrations are also provided to verify the proposed scheme.

• Current Optics and Photonics
• /
• v.5 no.2
• /
• pp.114-120
• /
• 2021

Free-space optical (FSO) communication is considered to be a potential solution to congestion in the radio-frequency spectrum and last-mile-access bottleneck issues in future cellular communication networks, such as 5G and beyond. However, FSO link performance may degrade significantly due to irradiance fluctuations and random temporal fluctuations from atmospheric turbulence. Therefore, in this work the main objective is to reduce the effect of the atmospheric turbulence by considering a multihop FSO communication system with amplify-and-forward relaying supported by a radio-frequency (RF) link, which form a hybrid FSO/RF communication system. The FSO link is assumed to follow the gamma-gamma fading model, which represents strong turbulence. Also, the RF link is modeled by a Rayleigh distribution. The performance of the considered system, in terms of the outage probability and average bit-error rate (BER), is investigated and analyzed under various weather conditions and pointing errors. Furthermore, the effect of the number of employed relay nodes on the performance of the system is investigated. The results indicate that the considered system reduces outage probability and average BER significantly, especially for low channel quality. Finally, the closed-form expressions derived in this work are compared to the results of Monte Carlo simulations, for verification.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.40 no.12
• /
• pp.2450-2460
• /
• 2015

Recently, as interest in IoT (Internet of Things) increase, research and standardization of a new protocol which reflects the characteristics of IoT has progressed. Among them, RPL(IPv6 for Low-Power Lossy Network) is a standardized routing protocol for IoT. RPL utilizes DIO (DODAG Information Object) messages which is flooded from the sink node to the whole network for path establish and maintenance. However, in large scale networks, not only a long time is required to propagate the DIO message to the whole networks but also a bottleneck effect around the sink node is occurred. Multi-sink based approaches which take advantage of reducing routing overhead and bottleneck effect are widely used to solve these problems. In this paper, we define 'unnecessary path update problems' that may arise when applying the RPL protocol to the multi sink based IoT networks and propose two methods namely Routing Metric based Path Update Decision method and Immediate Successor based Path Update Decision method for selective routing update.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09a
• /
• pp.213-215
• /
• 2003

Multimedia has to carry data of heterogeous types. Multicast communication techniques can supply the most appropriate infrastructures to such multimedia. Of many multicast protocols, the core based tree (CBT) protocol is the most concentrative studies are conducted on. The CBT places a core router at center of the shared tree and transfers data through the tore router. However, the CBT has two problems due to centralizing all network traffics into a core router. First it can raise bottleneck effect at a core router. Second, it is possible to make an additive processing overhead when core router is distant from receivers. To cope with the problems, this paper proposes an intelligent anycast routing protocol. The anycast routing attempts to distribute the centralized traffic into plural core routers by using a knowledge-based algorithm. The anycast routing estimates the traffic characteristics of multimedia data far each multicast source, and achieves effectively the distributing that places an appropriate core router to process the incoming traffic based on the traffic information in the event that request of receivers are raised. This method prevent the additional overhead to distribute traffic because an individual core router uses the information estimated to multicast sources connected to oneself and the traffic processing statistics shared with other core neuters.